The AWS connector is in Beta. Its scanners are verified against emulated AWS APIs; they have not yet been validated on a live customer account. Treat the list below as what the code reads, not as a track record.
Today the connector authenticates with an access key for a dedicated IAM user that you create, carrying a metadata-only policy (shown in full in the onboarding wizard) or the AWS-managed SecurityAudit policy. A cross-account role with an external ID is the pattern we want to move to; role assumption is not implemented yet. The key is encrypted at rest, and you can rotate or delete it in IAM at any time.
1. Compute
- EC2 — instances stopped for a long time, sustained high CPU, no detailed monitoring, public instances with no security group, legacy instance types.
- EBS — unencrypted volumes, unattached volumes, old snapshots, public snapshots.
- Lambda — deprecated runtimes, timeout and memory risks, functions that look abandoned.
- ECS / Fargate — services assigning public IPs, secrets in plain environment variables, missing log driver, no Container Insights.
- EKS — covered in the Kubernetes post.
2. Data
- S3 — public access, default encryption, versioning, lifecycle rules, access logging.
- RDS — instance availability, backups off or short retention, large single-AZ databases, automatic minor upgrades off, engine end-of-life.
- DynamoDB — encryption key type, point-in-time recovery, backup status, table-size growth between scans.
- Secrets Manager / SSM Parameter Store — names, rotation status and staleness, and plain-text parameters whose name looks like a credential. We never read a secret value.
3. Identity
IAM — access key age, MFA posture, root account usage and general credential hygiene. The classic finding: a console user with an old access key and no MFA. We do not compute effective permissions or privilege-escalation paths.
4. Networking & edge
Security groups open to 0.0.0.0/0 on a sensitive port (or every port), unused security groups, default-VPC use, VPC Flow Logs disabled. Load balancers, CloudFront distributions, API Gateway, WAF, and SQS/SNS each get configuration checks. Route 53: registered-domain expiry, CNAMEs that point at something deleted, health-check coverage.
5. Security services
- GuardDuty and Security Hub — whether each is enabled, and its active findings.
- AWS Health — open incidents in the regions you use, and scheduled changes such as EC2 instance retirements, RDS maintenance and runtime deprecations. AWS only offers this API on a Business- or Enterprise-level support plan; without one, the scan says so instead of failing.
- ACM — every certificate, with expiry alerts on the same 90/60/30/14/7/1/0-day ladder as everything else, plus validation status and renewal type.
- CloudTrail — no multi-region trail, log-file validation off, a public logs bucket, no CloudWatch Logs delivery, a trail switched off, no KMS encryption.
6. Backups and quotas
AWS Backup and RDS automated snapshots: each protected resource's newest recovery point against how often its plan runs, failed backup jobs in the last seven days, and databases nothing backs up. Service Quotas: usage against your applied quotas, flagged at 80% and 95%.
7. Cost
The AWS Cost Explorer page reads Cost Explorer when an account admin refreshes it, and caches the result. A nightly spend check — a day well above the week before it, and AWS Budgets reaching 75, 90 and 100% or forecast to pass them — is off unless it is switched on for your account, because each Cost Explorer request is billed to your AWS account (about a cent; roughly 30 cents a month per connection).
What the AWS connector does NOT do
- Never reads S3 object contents, database rows, DynamoDB items or secret values. It reads CloudTrail's management-event history to show who changed what, and keeps who, what and when, never the request body. The metadata-only policy contains no such permission, so it could not if it tried.
- Never writes. Every call is a read: List, Get and Describe calls, CloudTrail's LookupEvents, and IAM's GenerateCredentialReport, which only builds IAM's own credential report.
- Never moves between accounts. One credential per account, created by you.
GuardDuty, Security Hub and AWS Health findings — things AWS has already raised — are read every 5 minutes on a paid plan (every 15 during a trial). Exposure scans — public S3 access, world-open security groups, unauthenticated API Gateway routes, ECS tasks with public IPs — run every hour. Expiry and posture scans — ACM certificates, IAM access key age, Secrets Manager rotation, Route 53 domain expiry, backups and quotas — run once a night. An account administrator can ask for a fresh scan of a connection, at most once every 15 minutes per account; it runs with the next hourly scan, and while CloudMonitor is pre-launch nothing scans during the part of each day the service is powered down (see the Terms). Setup steps and the policy are on the Connectors page.